March 11, 2008
Coming Soon to Subways: An End to FAIL!
Hey, you know those awful flatscreen maps in the subways? The ones with the confusing station names and the pixelated Alcatel diagrams that look like a screengrab from an Atari 2600? Don't get too accustomed to them, because Muni and NextBus are working on something better. It'll take a few weeks -- at least -- but eventually, they'll be replaced by nicer, prettier, more accurate, comprehensible data. Hooray! That only took, what, a billion fucking years?
As an added bonus, the NextBus maps will finally be able to listen to Muni's own (antique) system, so the estimates will be more reliable. Of course, they'll still be located down by the track, rather than at the entrance to the station -- a security precaution, according to a Muni spokesman. Muni was apparently concerned that the expensive flatscreens would suffer vandalism, which sounds like a reasonable fear to us. (For example, we've seen how nicely our property is treated when it's in Hunters Point.)
If you're standing outside the subway, trying to decide whether to wait or walk, Muni recommends calling 311 to find out when the next train is coming (not to be confused with the useless 511, which still doesn't understand NextBus). But what about putting scrolling LED signs up near the subway entrances? They've got those hanging in bus stops all over the city, and they seem pretty vandal-resistant (except when they're not). Well, Muni says they might "consider various options." When our parents told us that, it usually meant "I'm busy enough already, kid" -- so we're guessing that having arrival times near the entrance is too much to hope for in the near future.
|
The NextBus displays currently in the kiosks are just never even close to right. A couple of days ago, I was in Civic Center station and the NextBus display said the next J would be in 22 minutes. I ignored it, went down to the platform, and was on a J in 3 minutes.
What gives? Is it like that for all the lines, or just the J?
Currently, Muni's system for tracking trains in the subways is literally an antique -- it barely works and it's hard for NextBus to talk to, but Muni can't afford to replace it. Luckily NextBus engineers are working on improvements (they'll be ready in a couple months, according to my source) that will improve the quality of the data that NextBus receives, making the maps much more accurate.
Every time I call 311, they tell me to IGNORE NextBus and only go by MUNI's printed timetables. This doesn't ever help, as the timetables only give an estimate of when the bus/metro should arrive. BUT not all buses (especially those above ground) turn on their tracker, and this causes a lot of anxiety. How about if MUNI invests in research to make those Star Trek-like transporters?
When I drive along Judah, I often pass a moving train. At the next stop, the display will say next train in 20 minutes, even though the train is right behind me. Just, another reason why fewer people ride MUNI:
http://www.examiner.com/a-1271248~Report__Public_transit_ridership_up_in_U_S___but_not_for_Muni.html
sorry, transporters would go against union rules since it would obsolete drivers.
Adding some of the gory details to MattyMatt's explanation, the Automatic Train Control System (ATCS, at least I think that's what it stands for) screen is updated every 5 seconds, and the displays on the station platform grab a screen shot a little less often than that. At least if you're lucky and don't get a broken image.
NextBus doesn't get that live feed though, which is what they're trying to fix now, it only receives an update from ATCS when a train arrives at a station. With only that to go on, since NextBus can't track trains when they're underground, it's just trying to make a best guess based on the schedule, historical data and whatever other voodoo they've baked into it.
I'm sure it'll be deployed right around the time TransLink is out of testing.
Uh? The NextBus display isn't usable on the large screens (and was apparently never licensed to MUNI for the big screens). There is one remaining rogue large NB display at West Portal, and a few of the metro stations had public facing smaller displays inside the station agent booths.
Just make the fucking subway run safely and reliably.
The Alcatel displays were never meant to be shown to customers.
Hence the charmingly functional, NASA Space Shuttle quality of the displays. I like them; they make me feel like an insider.
The perfect is the enemy of the good. I'm glad they rolled out version 1.0 last year; and I'm happier that they're continually improving the displays.
It could be so much worse. New York's MTA just canceled their entire bus arrival time project, for which they had hired incompetent moneysucking contractor Siemens at outrageous expense.
New Yorkers are envying us for choosing to employ startup NextBus instead of established bloodsucking contractors; I suppose it's just one of the public benefits of Silicon Valley culture.
While Nextbus does not work well on the metro, I love it on the streets. I was using my cell phone's web browser on a restaurant on 4th street, and it predicted the next bus each direction perfectly.
I measured NextBus on the 1 California line and found that it's less accurate than plain old guessing. The predicted arrival time of buses at Sacramento and Fillmore (both directions) wasn't even correlated with actual arrival times. NextBus is a mild joke on the surface and a bitterly unfunny joke in the subway.
We wouldn't even want NextBus if Muni service was at all reliable. We'd just go down to the corner secure in the knowledge that another train will arrive in the next six minutes. The only reason we want NextBus is so we can consult it to see if Muni is melting down today.
Did you report it to feedback -- at -- nextbus -- dot -- com? Because that's what good Muni watchers do when they see a NextBus issue.
I have found NextBus very helpful and mostly reliable for surface lines.
Today I needed to catch a 24-Divis to Pacific Heights. I fired up my browser and it said 10 minutes for my stop. I hurried out the door as it takes 6-7 minutes to walk to the stop. When I got to the stop, the bus was 4-5 blocks away.
Exactly. A functional MUNI system is what I'd rather see... a functional MUNI system would render NextBus a needless expenditure. Alas.
As explained on the Yahoo! Rescuemuni group, NextBus doesn't really account for missed runs, downed vehicles, misc delays or service interruptions, often doesn't account for two car trains(!!!), and it certainly doesn't account for vehicles that are too full to board. Which is to say that when a crush loaded 28 rumbles by and the driver refuses to let anyone board because there really isn't any room... NextBus still informs me that my bus is arriving.
NextBus is a nice TOY, but it doesn't make up for Natty Tatty Ford maintaining the status quo.
But will the new displays run pirated software?
My main complaint about NextBus is that when the train isn't moving at all, but the system still "sees" it, instead of the sign saying "No prediction available," it keeps re-estimating the time left, pissing everybody off when the number goes down then jumps back up. If a vehicle is at the same place for more than a few minutes, the system should automatically recognize that there's a problem, and not give a prediction until it starts moving again.
Or, if they want to be extra-helpful, it could say something like "train is stopped X minutes away" so that people know that there's something wrong, but they also know how long a wait they'll have after things get started up again.
theo: the ATCS screenshot has the advantage of being legible on a big screen. The NextBus display is not. The NB display doesn't even label the streets or the stations!
The Alcatel ATCS getup is antiquated... oh that's rich. The NYC metro's system is fifty odd years old and it handles higher volumes of traffic with comparative ease (case in point: the train I was on this morning was rendered immobile by the ATCS at West Portal and at Castro).
The Alacatel system is actually fairly high tech (moving block, bi-directional communication with the trains) and in many ways quite modern. Modern, however, does not equal reliable. Quite frankly, the people who actually believed the "ZOMG THERE ARE NOOOOO OS/2 PROGRAMMERS ANYMORE BECAUSE IT'S OLD AND SCARY" obviously a.) never used OS/2 and b.) aren't involved with embedded systems.
The issue is that MUNI tweaked the bidding process to favour a specific bidder that probably had close ties to our fedora festooned facade of a mayor. Unfortunately neither MUNI nor the contractors (Alcatel, Breda I'm looking at you) want to take responsibility for this crap and we, the riders, are left suffering.
One of the RM guys commented that New York is also moving to a moving block type ATCS, and that they're having problems. Ha. They haven't rioted in New York yet so the problems can't be THAT bad.
BART also uses a proprietary control system that predates ATCS, but it like NYC also has the advantage of being a completely closed system. If ATCS didn't have to contend with trains entering the system at multiple points, in random order, and at irregular intervals, it would be more reliable.
The Docklands Light Rail (DLR) in London is the only other system using this ATCS system, but the entire system is grade separated (which means, like BART, it never runs on the street with traffic or pedestrians) and is completely driverless.
Another problem is we really can't make any changes to ATCS on our own because they didn't make it easily configurable. This isn't a result of a rigged bidding process, but something Alcatel got in the contract that Muni and the City's lawyers missed.
You've likely noticed the T line does not appear on the ATCS display and that's because routing changes can only be made by the ATCS programmers, not by us or through the software preferences. They're going to charge us for that service of course, but it's going to be a while because the programmers are busy working on the DLR expansion project.
SFMTA engineers got a bit creative and hacked it so an unused destination so the overhead displays on the inbound platforms read "T Sunnydale", but that's as far as it can go until those programmers free up.
As for the statement about Mr. Ford maintaining the status quo, be sure to have a look at the Transit Effectiveness Project's proposed changes. Over the next couple months there will be a series of community meetings and you can give feedback through the website, after which the final service plan will be put together and go into effect next summer, though changes that require new infrastructure (like trolley bus wire) will take several years.
More immediately, schedulers are being hired now using Prop A funding to adjust schedules of the current system to match the reality of how long it actually takes to get from one end of a line to the other.
Here's an idea: turn off the ATCS and let the driver operate the train. How hard is that?
Don't think it isn't being considered, but there's more to it than just flipping the power switch.
Most important is keeping a safe stopping distance between trains. Operating on manual in a dark tunnel drivers would have to keep so long a distance and drive so slowly the tunnel would just grind to a halt. That's where block signals come in which divide the track into segments (blocks) and lets drivers know when it's safe to enter a block once the train ahead has cleared using traffic signals. ATCS can run the trains even closer together that normal block signaling because it centrally controls all the trains and can react faster than a human being ever could. But that system was left to degrade after ATCS was installed and the turnback after Embarcadero was designed and built for ATCS and never had block signals to begin with.
All the switches in the tunnel would also need to be redone to allow trains to flip them instead of ATCS handling it.
It's possible to remove ATCS, the issue is how to pay for it. The federal government will not give us any money because it was purchased on a 25 year replacement schedule, so we're on our own.
I understand the difficulty of ditching the ATCS but you are using the future tense "would just grind to a halt" to describe the situation we have in the present. Anybody lurching forward a meter at a time in the tunnel between Van Ness and Civic Center can tell you that the closed-loopedness of the ATCS does not translate into operational quality.
How were the Boeing cars controlled and without a fancy moving block signalling system how did they operate safely, on what headways, and carrying what passenger load? That's the only way to evaluate the success of the ATCS.
Perhaps the big bad meltdown in the late 90s has just permanently set me against the ATCS and I can't see the value of it no matter how well you argue it.
Ok, yeah it seems like more often than not train in the tunnel move about at speeds comparable to continental drift, but that's not all the time and on the rare occasion when everything works, trains do move along at a good clip with pretty even spacing.
Prior to ATCS, trains would be coupled, up to four cars long, before entering the tunnel and block signaling could be used because there were fewer total trains in the tunnel. With that setup each line could run on the exact same headway as now, but in the tunnel they operated as though there were only two lines. I didn't live here then, so I never got to experience this in action.
It wasn't without it's own problems, but even if the Breda trains we have now could be coupled into three- or four-car trains (they can't, some issue with them not being able to control a third car) the couplers were never meant to be used like that (since they had ATCS, they'd only have to couple or decouple a couple times a day at most) and now we have the Caltrain and Third Street extensions that would probably make it impossible.
I didn't mean to argue for ATCS, just provide more information about how it works and what it would take to get rid of it. I mentioned the DLR and BART to contrast them to Muni, which is not a closed system where everything can be tightly scheduled like they are. These are the constraints we're under, but if you have any ideas how we can get around them, please share.
Jamison: I'd been wondering why they stopped coupling unlike trains way back when. I thought it was that it was ridiculously inefficient, but then I couldn't see why they did it in the first place. Thanks!
Anyway, is there any reason (real reason, not Muni reason) they can't release detailed specs on what they would need for a replacement control system and let the bored masses of computer programmers (local and otherwise) have a stab at it? If nothing useful is produced, then no harm done, but if something does come out of it, it's a major win for everybody.
I've made the same suggestion about open sourcing the train control software, though no-one I've mentioned it to was in a position where they could make it happen.
I can see a really biggy why it could be hard to get the specs: the whole ATCS system is proprietary. It's not even an SFMTA issue, we depend on Alcatel (or formerly Alcatel, the division was sold off I head) not just for the software, but all the hardware it's running on. And that isn't just a couple PCs at Central Control (or as a friend put it: The West Portal Computer Museum) it's a whole network of equipment that monitors and talks to the trains.
I think we'd probably run into some weird security paranoia as well, but the more I think about it, the more open sourcing tunnel control seems like a good idea, especially if it could be designed more broadly so other cities could use it as well.
Ooh, good point. What I had in mind for "specs" was just a list of things that a hypothetical system would need to do, regardless of hardware. I figured it wouldn't be that hard to reverse-engineer and get a totally new software stack on the existing machines -- I'm pretty sure you can install Linux on a Tic-Tac these days. But I didn't take into account that there might be dodgy contract issues with the hardware itself, thus derailing (no pun intended) the whole thing.
Still, if people did come up with an open solution that would run on commodity hardware, it'd make the migration cost worthwhile in the long run.
I won't even touch the security thing, because it's such a massive pile of stupid.
(Side note: I'd even say that a system that could be used by all cities would be the best possible system for any particular city, since it by definition would be able to accommodate new lines and service changes without melting down.)
I've looked it up before and there is a standard spec for something called ATCS, but I don't know if it's at all related to our system.